IJCropSeed: An open-access tool for high-throughput analysis of crop seed radiographs

• Published in: Computers and electronics in agriculture 2020-08, Vol.175, p.105555, Article 105555
• Main Authors: Dantas de Medeiros, André, Junio da Silva, Laércio, Maria da Silva, José, Cunha Fernandes dos Santos Dias, Denise, Dias Pereira, Márcio
• Format: Article
• Language: English
• Subjects: Agricultural land; Biological properties; Crops; Destructive testing; Digital imaging; Freeware; Germination; High-throughput phenotyping; Image segmentation; ImageJ; Machine learning; Nondestructive testing; Optical properties; Physical properties; Radiographs; Seed quality; Seeds; Source code; X-ray
• ISSN: 0168-1699; 1872-7107
• DOI: 10.1016/j.compag.2020.105555

[Display omitted] •New free and easy-to-use tool for analysis of seed radiography of agricultural species.•Tissue integrity and seed morphometry can be assessed easily, rapidly, and accurately.•Seed viability and vigor can be indirectly assessed through automated analysis of seed radiographs. Optical technologies that are able to analyze physical properties of biological samples are increasingly drawing interest in modern agriculture. The use of X-rays for analysis of internal properties of agricultural products, such as seeds, has proven its worth in providing information regarding their quality in a non-destructive manner. However, visual evaluations of radiographic images are time-consuming, subjective, and highly prone to error. Therefore, it is necessary to develop methods that allow these analyses to be performed in an efficient and assertive manner. To that end, a free-access, open-source, and easy-to-use tool called IJCropSeed has been developed for high-throughput analysis of radiographic images of seeds from several agricultural crops. In addition, an experiment was conducted in which machine learning models were developed from the information obtained from the tool to predict the seed germination capacity and seedling vigor of Crambeabyssinica. The results showed that IJCropSeed had a high performance for the analysis of digital radiographic images of the 24 agricultural crops evaluated, with high speed and high precision of segmentation of the images. The use of parameters obtained with the tool, in combination with the machine learning models, proved to be highly efficient in classifying the quality of C. abyssinica seeds. It is a non-destructive and highly effective method.